Schrödinger

The lowest AI centrality grade in this category, and it is not a criticism so much as an accurate reading of what is being bought. The differentiating asset is physics, not machine learning: FEP+ free energy perturbation, Glide docking, Desmond molecular dynamics and Jaguar quantum chemistry are simulation engines built on more than 30 years of R&D, and they run without any model.

Machine learning is real but layered on top, principally through the LiveDesign ML module for training and deploying property prediction models and through ML triage ahead of expensive physics calculations. The company states its position openly, that machine learning is most useful paired with physics based methods rather than treated as a substitute, and that refusal to inflate the AI claim is itself a candour benchmark in a category where the incentive runs the other way. Buyers should read this grade as a description of the mechanism, not of the quality.

The platform is a design environment for human chemists rather than an autonomous designer, which is the correct posture for a licensed tool and structurally limits the autonomy question. LiveDesign is explicitly collaborative, presenting physics based and machine learning calculations alongside experimental data so that medicinal chemists prioritize compounds before committing to synthesis. The decision to make a molecule stays with the customer's scientists.

Held at B rather than A because no published disclosure was located on confidence thresholds, prediction reliability bands, or guidance on when a computed result should not be relied upon, which is the disclosure that would let a buyer calibrate trust in the output.

Among the strongest in the index on this axis. The methods are published and independently benchmarked across a large peer reviewed literature rather than described in marketing terms, product documentation and a knowledge base are publicly reachable, a Python API is published, and the company runs free and certificated courses that teach the underlying theory including a dedicated course on free energy calculations with FEP+.

Named methodological advances are described specifically rather than generically, for example FEP+ Pose Builder as an integrated pipeline feature. A customer can read how the calculation works, reproduce it, and disagree with it.

The clinical form of this axis does not reach this product and the analogous form does, so it is graded on the latter rather than marked not applicable. The platform operates on molecular structures, simulation output and experimental chemistry data with no patient records in the workflow, which bounds the axis structurally and honestly: there is no protected health information for a chain to carry, and that is a fact about the product rather than a claim about its controls.

What replaces it is a genuinely sensitive asset. A customer's proprietary compound structures and target information are the most commercially valuable material that organisation holds, and a competitor learning which targets a company is pursuing is a material harm even though no patient is involved. That is addressed through deployment architecture and contract rather than through anything published, so a prospective buyer can establish nothing before entering a sales process.

Nothing is enumerated: no hosting provider named, no sub processor list, no retention position, and no statement of whether customer submitted structures or results contribute to improving general models, which is the same question this index asks of clinical vendors and which matters equally here. Ask which deployment options exist, who hosts them, for a sub processor list, and for an explicit statement that customer structures do not train shared models.

Evidence exists at both the platform and the asset level, which is rare here. At asset level, the Nimbus originated TYK2 inhibitor zasocitinib (TAK-279), designed using the platform, advanced into Phase III, and the company's own SGR-1505 MALT1 inhibitor received FDA Fast Track designation for relapsed or refractory Waldenstrom macroglobulinemia with initial Phase 1 data reported at the EHA Annual Congress.

At platform level the installed base is the evidence: the software is used across the top 20 pharmaceutical companies by revenue and in more than 70 countries, with published customer case studies including Morphic Therapeutic's alpha4beta7 integrin inhibitor. Caveat for buyers: widely cited figures for cycle time or synthesis reduction come from case studies and vendor materials rather than controlled comparisons, so treat magnitude claims as illustrative.

Not applicable in the provider sense and rated accordingly rather than penalized. The platform operates on molecular structures, simulation output and experimental chemistry data, with no patient records in the workflow. The analogous stewardship question for this buyer is protection of proprietary compound structures and target information, which is addressed through deployment architecture and contract rather than through a PHI policy.

Not applicable. Customers are pharmaceutical, biotechnology, industrial and academic research organizations licensing simulation software, not covered entities handing over protected health information, so no business associate relationship arises.

No SOC 2, ISO 27001 or equivalent attestation was located and no trust centre was found. But the company is a Nasdaq listed registrant, so a description of its cybersecurity programme exists in a filed and legally attested document, and that is where the substance is.

Its annual report carries the required cybersecurity disclosure and it is not boilerplate. The company describes processes for assessing, identifying and managing cybersecurity risk built into the information technology function, and enumerates physical, procedural and technical safeguards, response plans, regular system testing, incident simulations and routine policy review. It states that it engages external consultants, computer security firms and risk management advisors, and that it reviews the internal risk oversight programmes of third party service providers before engaging them. Governance is specific: the audit committee of the board has direct oversight of cybersecurity risk with periodic management updates and notification between updates of significant new threats or incidents, and a vice president of information security leads the function.

Graded B on that basis, and the distinction matters. This is a described programme with named governance and an accountable executive, published under a filing obligation. It is not an independent examination of controls, which is what an attestation provides, and a buyer should not treat the two as equivalent.

The commercial point stands regardless of PHI being out of scope: customers place proprietary chemistry into the environment and that chemistry is often the most valuable intellectual property the organisation holds. The mitigation available here remains architectural rather than attested, since the software can be downloaded and run on customer controlled infrastructure, which removes the question instead of answering it.

Correctly structured for this category: the software platform is not a regulated medical device and is not presented as one, while regulatory engagement is real at the asset level. SGR-1505 holds FDA Fast Track designation for relapsed or refractory Waldenstrom macroglobulinemia and is in clinical development, SGR-3515 is a disclosed Wee1/Myt1 program, and a platform designed molecule from a partner program reached Phase III. Buyers should note that Fast Track designation is a development and review process mechanism, not an approval or a finding of efficacy.

The strongest disclosure in this category on technical validity, though the company does not present any of it as governance.

The domain relevant question here is chemical rather than demographic: whether the methods perform unevenly across target classes and chemical space, and whether the company says where they degrade. Four things answer it directly. The company published a peer reviewed assessment of its free energy method's accuracy in a Nature portfolio journal, and released the assembled benchmark data set openly on a public versioned repository so others can reproduce and extend it. Its published work reports that accuracy varies between targets and chemical series rather than quoting a single headline figure. It publishes systematic failure analysis: in the protein binding work it examined every case with an error above 2 kcal per mole, quantified them at roughly 9 percent of the data set, and categorised the sources into sampling failures and force field limitations, explicitly identifying cases falling beyond the method's domain. And it discusses domain of applicability as an explicit topic rather than implying universal reliability.

An independent large pharmaceutical company also published its own benchmark of the method across multiple targets and hundreds of ligands. Submitting a product to third party evaluation that it can lose is rare in this category and the index credits it wherever it occurs.

Upgraded from B. It was previously held down for consistency with a peer that lacks a formal framework, a model card and an intended use statement. That was the wrong test. Consistency means applying the same criteria to every vendor, not arriving at the same grade, and on the criteria this record is materially more complete than the peer it was matched to: decomposition by target and by chemical series against that peer's single denominator, an openly released benchmark set, a quantified and categorised failure taxonomy, and independent third party benchmarking. The question this axis asks is substantively answered.

What is still absent is worth stating plainly, because A is not a statement that nothing is missing. There is no model card, no formal governance framework, and no applicability statement in the product documentation where a buyer would actually look. The material to write one already exists in the company's own publications, and publishing it would close the gap between what this company knows about its methods and what a customer can find.

The method is published, independently benchmarked across a large peer reviewed literature, and taught. That last part is unusual enough to name as the distinguishing feature. Product documentation and a knowledge base are publicly reachable, a programming interface is published, and the company runs free and certificated courses teaching the underlying theory, including a dedicated course on the free energy methods its headline capability rests on.

Teaching customers the physics well enough to check your calculation is the opposite of the disclosure posture this index usually encounters, where a vendor's advantage depends on the customer not being able to evaluate the claim. The practical effect is that a customer can read how the calculation works, reproduce it and disagree with it, and disagreement is grounded in a shared literature rather than in a dispute about a proprietary score.

Named methodological advances are described specifically rather than generically. Held below the top grade because nothing attaches commercially: no warranty, indemnity, service level or remediation commitment was located, and published benchmark performance for a method is not the same as a committed error characteristic for a customer's own system. Ask what accuracy the deployed configuration achieves on the target classes you work on, and where the method is known to fail.

Not applicable in the provider sense. There is no EHR touchpoint and no clinical workflow surface. The domain equivalent is research informatics interoperability, and here the story is genuinely strong: LiveDesign is positioned as an enterprise informatics layer unifying physics calculations, machine learning predictions and experimental data across discovery teams, and in January 2026 Lilly TuneLab workflows were made available inside LiveDesign with LiveDesign named a priority interface for participating biotech companies.

Better disclosed than most of this category because the product is genuinely licensed software rather than a partnership. A public release download portal and documented local installation mean the platform can run on customer controlled infrastructure, so proprietary structures need not leave the organization, and cloud scale execution is available for large virtual screens and ensemble calculations. Held at B because no explicit data residency terms, tenancy model or regional hosting commitments were located in public materials.

The Artrya precedent applies: a public listing forces disclosure that private peers can decline. As a Nasdaq registrant (SDGR) the company files software revenue, customer metrics, pipeline spend and risk factors on a regular cadence, maintains a public investor relations site with press releases and presentations, and publishes its end user license agreement, terms of use and other policies openly.

There is still no published rate card for the software, and enterprise licensing is quoted through sales, so this is transparency of the business rather than of the price. Even so it is the most inspectable commercial surface in this category.

The broadest coverage in this category by a wide margin. Within life science the platform spans small molecule discovery (structure prediction and target enablement, hit discovery, hit to lead and lead optimization, formulation) and mixed modalities including antibody design, peptide discovery, enzyme engineering, bifunctional degraders and TCR modeling.

A parallel materials science line covers polymers, organic electronics, energy storage, thin film processing, catalysis and metals and ceramics. Users span the top 20 pharmaceutical companies, biotech, academia and government laboratories across more than 70 countries.